Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
AI Technology

Microsoft CEO Nadella Taunts AI Rivals: Even With All the Hoopla, GPT-4 Remains the Best (techcrunch.com) 59

Posted by msmash from the how-about-that dept.
An anonymous reader shares a report: Microsoft's prescient bets and aggressive investments in AI have propelled the software giant to become the world's most valuable company. Yet Satya Nadella, its typically reserved chief executive, couldn't resist landing a gloved jab at the rest of the industry. "We have the best model today ... even with all the hoopla, one year after, GPT4 is better," Nadella said at a company event in Mumbai on Wednesday. "We are waiting for the competition to arrive. It will arrive, I'm sure, but the fact [is] that we have the most leading LLM out there."
This discussion has been archived. No new comments can be posted.

Microsoft CEO Nadella Taunts AI Rivals: Even With All the Hoopla, GPT-4 Remains the Best More

Microsoft CEO Nadella Taunts AI Rivals: Even With All the Hoopla, GPT-4 Remains the Best

Comments Filter:

  • This ain't a sprint, it's a marathon (Score:5, Insightful)

    by Opportunist ( 166417 ) on Wednesday February 07, 2024 @08:17AM (#64221644)

    Remember when Webcrawler was the biggest search engine on the net? When ICQ was the messenger of choice? When IE dominated the web browser world and everyone had to bend to its whims and flaws? When people connected via SixDegrees?

    That were the early days of web search, of instant messaging of browsing, of social media. Today, either of these markets are a few magnitudes bigger, and the technologies mentioned are left in the dust by Google Search, by WhatsApp, by Chrome, by Facebook. Nobody cares anymore who was the top dog back when those mountains were molehills.

  • CEO says his product is the best (Score:5, Insightful)

    by thegarbz ( 1787294 ) on Wednesday February 07, 2024 @08:53AM (#64221714)

    Or at least the one he has had his company invest in. News at 11.

    This isn't news, this is marketing.

    • And anytime a competitor emerges they'll buy them up too.

      • Not if the competitor is big enough (they can't buy Google or Facebook) or if the competitor is protected by a license like GPL (buying would be useless, a fork will continue).

    • They don't want to buy the massive copyright liability. But they will happily rent it.

  • Speaking of high tech competition, there was a TV show called "Halt and Catch Fire" on AMC a few years all about the early days of the tech revolution in the late 70s, early 80s.

    If any of you Slashdotters haven't seen it, it's four seasons of the best nerd-ish drama you'll ever see. It's not about any of the big names or big companies you've ever heard of, or even the technology exactly, though that plays a big part.

    The writing is utterly fantastic, the acting is great, and the dialog is well written. Hop

  • Death by filters (Score:3)

    by Craefter ( 71540 ) on Wednesday February 07, 2024 @09:42AM (#64221842)

    My prediction is that ChatGTP will be a political correct and safe enterprise service because OpenAI has a heavy focus on sanitizing ChatGTP's output. This will not satisfy people who do not want to live in a bubble, who will seek alternatives like Grok. In the end all LLMs will be on par and people will subscribe to the one which fits ones personality.

    • ChatGPT may become entrenched due to the network effect - all the external things it can talk to. Its bing integration is crucial, so it isn't limited to what is built into its model weights. There isn't really anybody else with a first-rate model and search engine except google, but they have to push search predominantly because they are making so much money from ads. Just asking a question and getting an answer would kill their cash cow. (Gradually I suppose there will be a coming together and 'chat' w

    • Re: (Score:3)

      by Rei ( 128717 )

      who will seek alternatives like Grok.

      Lol [google.is].

      Anything made by a company is going to inherently be more limited than those made by individuals, as companies are under more pressure from legal liability risks.

      • Considering Musk will deliberately neuter anything his pile of code spits out, you can be assured it will be in the trash heap in no time when compared to other AI generators.

  • AI (Score:4, Informative)

    by stealth_finger ( 1809752 ) on Wednesday February 07, 2024 @09:45AM (#64221848)
    Stop calling it AI for fucks sake. It's nothing more than predictive text.

    • Adding "and show your steps" will elicit a reasoning chain from those AI's. And it can be demonstrated that it's not just regurgitation. If "predictive text" can demonstrate both reasoning and originality, than "It's just predictive text" is similar to saying "high marks on exams aren't indicative of true intelligence". It might be true, but faking it is a harder problem than just doing it.

      • Just because it has you convinced doesn't mean it's intelligent. They've never made any secret out of the fact it is predictive text. Extremely good predictive text I'll grant you but it's still predictive text.

        • What kind of definition do you have for the word "intelligence" if it doesn't include reasoning and originality?

          I didn't say that AI's have a high level of intelligence. They're obviously often pretty stupid. But they have intelligence by any reasonable definition of the word, IMO.

          • That's the problem right there. They are not engaging in any reasoning. They are carrying out statistical inference.

            Reasoning is not the same thing as statistical inference.

          • What kind of definition do you have for the word "intelligence" if it doesn't include reasoning and originality?

            Well, that's the thing. I would argue they are doing either of those. Maybe originality by the strictest of definitions but they definitely don't reason despite how well they pretend to when prompted.

    • I just finished an AI course from IBM and they referred to it as "Augmented Intelligence". Regardless, there needs to be some word to describe this new technology, which is capable of far more than just predictive text.

      • lol It's not the artificial part of the name I take umbrage with. Maybe call them I dunno, large language models? Or just chatbot is fine.

        • LLM is just one subset of AI.

          • Exactly. So call it was it is instead of a generic term that already had a fairly fixed definition in most people's mind and it's not five rounds of conversation before it goes completely off the rails.

            • I don't know why I'm bothering, but RTFA. He uses the word LLM appropriately and also the word AI correctly. English is hard, I know.

      • Re: (Score:2)

        by ceoyoyo ( 59147 )

        There is. AI. Predictive text is also an example, but a less impressive one.

        Artificial intelligence is term that was explicitly coined to describe an effort to invent algorithms that would let (newly invented) electronic computers solve problems that appeared to be easy for human intelligence but very difficult for conventional algorithms.

        The fact that people like the OP are threatened by the words and prefer to make mystical claims about what they suppose they mean is irrelevant.

        • Where did you get I was threatened from? At best annoyed at the use of a general term to describe a specific thing. Tell you what. Get an llm to do anything other than spit out text. Combine it real time visual processing and speech recognition, give it a way to do new things that don't involve millions of hours training an a supercomputer and then give it some way to actually interact with the world or people and then we can talk about calling it an AI.

          • Re: (Score:2)

            by ceoyoyo ( 59147 )

            AI is a specialized, technical term.

            I suppose your annoyance could be ignorance of that. Usually though, deep down, it's being threatened that your intelligence isn't as unique and special as you thought. I'm still leaning toward the latter, given both your original and this post. Particularly since you still insist on making up your own definition when the actual one has been given.

    • Re: (Score:3)

      by Rei ( 128717 )

      Predict the next word in the following sentences:

      "The odds that Russia will win the war in Ukraine are..."
      "The odds that there will be a major terrorist attack in Ireland carried out by a national of Ghana over the next 10 years is..."
      "In a cage fight between Joe Biden and Justin Bieber, the probable winner would be..."
      "In short, to answer the question [insert complex question here], the answer is: ..."

      Can you just put any word there? No. Because mere "text" is a reflection on the broader world, and to be a

      • Re: (Score:1)

        by Anonymous Coward

        AI is not even remotely a new term, and it's funny how people only started taking issue with it recently, when there were huge advances in the field.

        I remember people were arguing about the term years before the modern generative AI boom. Taking issue with it isn't new either.

      • Ok cool, but none of that changes the fact that it is predictive text and not AI. Granted it's orders of magnitude more effective and accurate than what's on your phone but that's still what it is. They have never tried to deny or hide that. I see you around here all the time and you seem like a smart guy so I'm sure you know all about tokens and how these things actually work (as well as anyone can at least). Yeah they can pass the turing test in certain circumstances now but I don't think anyone would say

    • Not this again. There is no clear consensus definition, partly because we don't even know how the human brain fully works. Plus, it should be defined for the job it does, not how it's implemented. Yet there is no clear-cut definition of intelligence itself, I've been in several dozens of debates on such. Trying to define it via Boolean features (has vs. not has) is a fool's errand. And if left a continuum, nobody agrees on where to draw lines to say if a gizmo is "AI". Give up AI definition bashing, it's a

      • Well here's a start: Can do more than one thing and adapt to unknown inputs without a million hours of training on supercomputer cluster.

    • AI is a field of scientific inquiry -like mathematics.

      Your lack of understanding does not mean that division is not mathematics -nor that LLMs are not AI.

      • Right, but when you are talking about math you have to be specific to type you are talking about other things get confusing real fast.

        It's like calling an ingredient a meal because it's part of the full thing but on its own, an onion, for example, is not dinner.
        • *otherwise
        • AI is a field to study/understand/simulate/define intelligence using machines. LLMs are part of AI research. They don't have to be "intellegent" to be part of AI research. This is just a historical and useful name given to this field of study.

    • It is, and it isn't

      Remember when the first chess-playing computers came out? It was thought back then that any computer that could beat the world grand master in chess, would be super-intelligent--would possess AI. That computer has been built, the world grand master is no longer a match for the best chess programs. Are these computers "super-intelligent" or some kind of AI? No, not really, but they are *really* good at chess.

      LLMs are kind of similar. They do hit the mark when it comes to being able to "con

  • Satya, you basically bought the thing you're touting. Being proud of it is one thing - thumping your chest over it, as though you or your company contributed to it in any significant way beyond funding or providing Azure access, is quite another. You're just being tacky and juvenile.

  • Hold my beer...
  • Buying it is the way to get ahead!

  • "The best" what?

Slashdot Top Deals

Make sure your code does nothing gracefully.

Close